Postgresql：使用窗口函数从日期范围中选择最大值

我试图从过去两年内选择最大值

例如，如果我有一个如下表：

|person_id | pass_or_fail | timestamp|
|----------|--------------|-----------|
|1234      | 1            | 1990-01-01|
|1234      | 0            | 1995-01-01|
|1234      | NULL         | 1995-12-12|
|6789      | 0            | 1990-01-01|
|6789      | 0            | 1991-01-01|
|6789      | 1            | 1995-01-01|
|6789      | 1            | 1996-01-01|
|6789      | 0            | 1997-01-01|
|6789      | NULL         | 1997-03-03|

我想从我的查询中获得以下信息：

person_id |highest_grade_from_past_two_years | pass_or_fail | timestamp
1234      |1                                 | 1            | 1990-01-01
1234      |0                                 | 0            | 1995-01-01
1234      |0                                 | NULL         | 1995-12-12
6789      |0                                 | 0            | 1990-01-01
6789      |0                                 | 0            | 1991-01-01
6789      |1                                 | 1            | 1995-01-01
6789      |1                                 | 1            | 1996-01-01
6789      |1                                 | 0            | 1997-01-01
6789      |1                                 | NULL         | 1997-03-03

---

我如何编写窗口函数来给出这个结果？

我不知道如何使用窗口函数编写这个查询。不带窗口的版本：

SELECT t1.person_id,
       max(t2.pass_or_fail) AS highest_grade_from_past_two_years,
       t1.pass_or_fail,
       t1.timestamp
FROM t AS t1
     JOIN t AS t2 ON (
          t1.person_id = t2.person_id
          AND t2.timestamp < t1.timestamp + '2 year'::interval
          AND t2.timestamp <= t1.timestamp
     )
GROUP BY 1, 3, 4

---

我不知道如何使用窗口函数编写此查询。不带窗口的版本：

SELECT t1.person_id,
       max(t2.pass_or_fail) AS highest_grade_from_past_two_years,
       t1.pass_or_fail,
       t1.timestamp
FROM t AS t1
     JOIN t AS t2 ON (
          t1.person_id = t2.person_id
          AND t2.timestamp < t1.timestamp + '2 year'::interval
          AND t2.timestamp <= t1.timestamp
     )
GROUP BY 1, 3, 4

---

通过与实际第1年（如果存在）合并

select p1.person_id, p1.pass_or_fail, p1.timestamp,
       case when coalesce(p1.pass_or_fail,0) > coalesce(p2.pass_or_fail,0) 
            then coalesce(p1.pass_or_fail,0) else coalesce(p2.pass_or_fail,0) end as highest_grade_from_past_two_years
from person p1
     left join person p2
     on p1.person_id = p2.person_id
        and extract(year from p2.timestamp) = extract(year from p1.timestamp) - 1
order by p1.person_id, p1.timestamp

)

---

勾选此处：

与实际年份-1（如果存在）连接

select p1.person_id, p1.pass_or_fail, p1.timestamp,
       case when coalesce(p1.pass_or_fail,0) > coalesce(p2.pass_or_fail,0) 
            then coalesce(p1.pass_or_fail,0) else coalesce(p2.pass_or_fail,0) end as highest_grade_from_past_two_years
from person p1
     left join person p2
     on p1.person_id = p2.person_id
        and extract(year from p2.timestamp) = extract(year from p1.timestamp) - 1
order by p1.person_id, p1.timestamp

)

---

检查这里：

我看不到使用窗口功能的明显方法。相关子查询或横向联接将起作用：

select t.*,
       (select max(t2.pass_or_fail)
        from t t2
        where t2.person_id = t.person_id and
              t2.timestamp <= t.timestamp and
              t2.timestamp >= t.timestamp - interval '2 year'
       ) as highest_grade_from_past_two_years
from t;

---

我认为使用窗口函数可以得到的最接近的范围规范。但是，Postgres不支持前面的范围规范。

我看不到使用窗口函数的明显方法。相关子查询或横向联接将起作用：

select t.*,
       (select max(t2.pass_or_fail)
        from t t2
        where t2.person_id = t.person_id and
              t2.timestamp <= t.timestamp and
              t2.timestamp >= t.timestamp - interval '2 year'
       ) as highest_grade_from_past_two_years
from t;

---

我认为使用窗口函数可以得到的最接近的范围规范。但是，Postgres不支持前面的范围规范。

上面的第二个表是我希望查询结果的样子。我只使用一个表两个表都是相同的，我使用它是为了避免在select语句中使用子查询-它分别为每一行运行上面的第二个表就是我希望查询结果的样子。我只处理一个表这两个表都是相同的，我使用它是为了避免在select语句中使用子查询-它以每行分隔符的形式运行-将子查询放在select中不是很好-它以每行分隔符的形式运行。@RomanTkachuk。我认为您对SQL的理解不如我。这里可以使用相关子查询。如果在tperson\u id、timestamp、pass\u或fail上有一个索引，它将运行得更快。将子查询放在select中不是很好，它会为每一行单独运行。@RomanTkachuk。我认为您对SQL的理解不如我。这里可以使用相关子查询。如果在tperson\u id、timestamp、pass\u或fail上有索引，它将运行得更快。